The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Mixed-Signal Design >> How we select PRBS sequence for BER test in SerDes
https://designers-guide.org/forum/YaBB.pl?num=1112691037

Message started by raymond_luo2003 on Apr 5^th, 2005, 1:50am

Title: How we select PRBS sequence for BER test in SerDes
Post by raymond_luo2003 on Apr 5^th, 2005, 1:50am

Dear all,

I read many papers regarding on the BER test of SerDes link. Different papers use different PRBS sequence 2E7-1; 2E11-1; 2E15-1; 2E23-1 to test their BER. For a same link, different PRBS will result in different BER, e.g. 10E-12 or10E-7.

Does anyone can help me understand which PRBS sequence should use in a typical link, say fibre or USB SerDes? Why should I choose one instead of another one? What’s kind of result should I expect if I select a particular PRBS? The rule for choosing particular PRBS is related to eye-diagram of transmitter or receiver? Is that related to the bandwidth of PLL or CDR used in SerDes?

Thanks in advance!

Raymond

Title: Re: How we select PRBS sequence for BER test in Se
Post by Paul on Apr 5^th, 2005, 1:27pm

Hi Raymond,

The main difference between these PRBS sequences is the repetition interval (number of bits until the same sequence repeats) and the runlength (maximum number of consecutive identical digits, also called CID). CID defines the memory requirements of your PLL. For most short-distance applications, you have 8b/10b encoding or 64b/66b (e.g. 10GbE) which defines the maximum runlength. Using a sequence with a lare CID to test a short-distance protocol is to stringent.

The patterns to be applied, as well as the jitter to be added to the input signal, are described in this document:
ftp://ftp.t11.org/t11/pub/fc/mjsq/04-101v4.pdf

It is true that several papers show BER measurements with a pure input signal (no jitter applied), but achieving BER < 1E-12 in these conditions is not sufficient to make your design work in a real life situation.

You should also check your application spec (I'm sure the USB spec gives some references for BER measurement), most short-distance standards refer to the above mentioned Fibre Channel document. SONET and other long-haul systems have much tougher specs.

Best,

Paul

Title: Re: How we select PRBS sequence for BER test in Se
Post by Paul on Apr 6^th, 2005, 2:23am

Raymond,

one more paper of interest about SerDes jitter tolerance measurements:

Multi-gigabit-rate clock and data recovery based on blind oversampling
Jaeha Kim; Deog-Kyoon Jeong;
Communications Magazine, IEEE
Volume 41, Issue 12, Dec. 2003 Page(s):68 - 74

Paul

Title: Re: How we select PRBS sequence for BER test in Se
Post by DReynolds on Apr 7^th, 2005, 6:02am

Another addition to Paul's comment: more is always better. I have seen 8b/10b systems (max run length =5) tested with much longer prbs than 2^^7 -1 because it gives an indication of margin . If your cdr can tolerate the longer run then stressing it with the longer sequence will help you find the limits of your performance better than the shorter sequence.

David Reynolds

Title: Re: How we select PRBS sequence for BER test in Se
Post by raymond_luo2003 on Apr 27^th, 2005, 2:44am

Thanks Paul and DReynolds

I didn't realized that such good reply is in place since I didn't ask for notification.

Thanks indeed